Performing Texture Segmentation Using Gabor Filters

Image segmentation can help you select regions of interest in an image. Image Processing Toolbox™ offers a
variety of techniques for image segmentation.
This example shows how to use texture segmentation to identify regions of an image based on their tex-
ture—in this case, to segment the dog from the bathroom floor. The segmentation is visually obvious
because of the difference in texture between the regular, periodic pattern of the bathroom floor, and the
regular, smooth texture of the dog’s fur.
From experimentation, it is known that Gabor filters are a reasonable model of simple cells in the mamma-
lian vision system. Because of this, Gabor filters are thought to be a good model of how humans distin-
guish texture, and are, therefore, a good model to use when designing algorithms to recognize texture.
This example uses the basic approach described in “Unsupervised Texture Segmentation Using Gabor
Filters” (A. K. Jain and F. Farrokhnia, 1991) to perform texture segmentation.

Reading and Displaying the Input Image

Suppose you have an image of a dog, with the filename kobi.png. You can get started by reading the image
into MATLAB® and displaying it. You can use the function imresize to shrink the image to make the
example run more quickly.

A = imread('kobi.png');
A = imresize(A,0.25);
Agray = rgb2gray(A);
figure
imshow(A)
Designing an Array of Gabor Filters
You can design an array of Gabor filters that are tuned to different frequencies and orientations. The set of
frequencies and orientations is designed to localize different, roughly orthogonal, subsets of frequency and
orientation information in the input image. To start, try regularly sampling orientations between 0 and 150
degrees in steps of 30 degrees, and sampling wavelength in increasing powers of two starting from 4/sqrt(2)
up to the hypotenuse length of the input image. These combinations of frequency and orientation are taken
from the paper cited in the introduction.

imageSize = size(A);
numRows = imageSize(1);
numCols = imageSize(2);

wavelengthMin = 4/sqrt(2);
wavelengthMax = hypot(numRows,numCols);
n = floor(log2(wavelengthMax/wavelengthMin));
wavelength = 2.^(0:(n-2)) * wavelengthMin;

deltaTheta = 45;
orientation = 0:deltaTheta:(180-deltaTheta);

g = gabor(wavelength,orientation);

Then, you can extract gabor magnitude features from the source image. It is common to work with the
magnitude response of each Gabor filter. Gabor magnitude response is also sometimes referred to as
“Gabor energy.” Each MxN Gabor magnitude output image in gabormag(:,:,ind) is the output of the
corresponding Gabor filter g(ind).
gabormag = imgaborfilt(Agray,g);

Postprocessing the Gabor Magnitude Images into Gabor Features

To use Gabor magnitude responses as features for use in classification, some postprocessing is required.
This postprocessing includes performing Gaussian smoothing, adding additional spatial information to
the feature set, reshaping the feature set to the form expected by the pca and kmeans functions, and nor-
malizing the feature information to a common variance and mean.
Each Gabor magnitude image contains some local variations, even within well-segmented regions of con-
stant texture. These local variations will throw off the segmentation. You can compensate for these varia-
tions using simple Gaussian low-pass filtering to smooth the Gabor magnitude information. To do this,
choose a sigma that is matched to the Gabor filter that extracted each feature. You can introduce a smooth-
ing term K that controls how much smoothing is applied to the Gabor magnitude responses.
for i = 1:length(g)
sigma = 0.5*g(i).Wavelength;
K = 3;
gabormag(:,:,i) = imgaussfilt(gabormag(:,:,i),K*sigma);
end

When constructing Gabor feature sets for classification, it is useful to add a map of spatial location infor-
mation in both X and Y. This additional information allows the classifier to prefer groupings that are close
together spatially.

X = 1:numCols;
Y = 1:numRows;
[X,Y] = meshgrid(X,Y);
featureSet = cat(3,gabormag,X);
featureSet = cat(3,featureSet,Y);

You can then reshape data into a matrix X of the form expected by the kmeans function. Each pixel in the
image grid is a separate data point, and each plane in the variable featureSet is a separate feature. In
this example, there is a separate feature for each filter in the Gabor filter bank, plus two additional features
from the spatial information that was added in the previous step. In total, there are 24 Gabor features and
two spatial features for each pixel in the input image.

numPoints = numRows*numCols;
X = reshape(featureSet,numRows*numCols,[]);

Next, you can normalize the features to be zero mean, unit variance.

X = bsxfun(@minus, X, mean(X));
X = bsxfun(@rdivide,X,std(X));

You can then visualize the feature set. To get a sense of what the Gabor magnitude features look like, you
can use principal component analysis to move from a 26D representation of each pixel in the input image
into a 1D intensity value for each pixel.
coeff = pca(X);
feature2DImage = reshape(X*coeff(:,1),numRows,numCols);
figure
imshow(feature2DImage,[])

It is apparent in this visualization that there is sufficient variance in the Gabor feature information to
obtain a good segmentation for this image. The dog is very dark compared with the floor because of the
texture differences between the dog and the floor.

Classifying Gabor Texture Features Using kmeans

You can avoid local minima when searching for means that minimize objective function by repeating
k-means clustering five times. The only prior information assumed in this example is how many distinct
regions of texture are present in the image being segmented. In this case, there are two distinct regions.
This part of the example uses Statistics and Machine Learning Toolbox™.

L = kmeans(X,2,'Replicates',5);

After performing the k-means clustering, you can visualize the segmentation using label2rgb.

L = reshape(L,[numRows numCols]);
figure
imshow(label2rgb(L))
You can then visualize the segmented image using imshowpair, and examine the foreground and back-
ground images that result from the mask BW that is associated with the label matrix L.

Aseg1 = zeros(size(A),'like',A);
Aseg2 = zeros(size(A),'like',A);
BW = L == 2;
BW = repmat(BW,[1 1 3]);
Aseg1(BW) = A(BW);
Aseg2(~BW) = A(~BW);
figure
imshowpair(Aseg1,Aseg2,'montage');
 Learn More About Image Processing
• Computer Vision with MATLAB 43:32
• Developing and Deploying Sonar and Echosounder Data Analysis Software (technical article)
• Creating Computer Vision and Machine Learning Algorithms that Can Analyze Works of Art
(technical article)
• Image Processing Toolbox (product trial)

© 2016 The MathWorks, Inc. MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See mathworks.com/trademarks for a list of additional trademarks.
Other product or brand names may be trademarks or registered trademarks of their respective holders.

mathworks.com

93008v00 05/16